CN103324921A - Mobile identification method based on inner finger creases and mobile identification equipment thereof - Google Patents
Mobile identification method based on inner finger creases and mobile identification equipment thereof Download PDFInfo
- Publication number
- CN103324921A CN103324921A CN2013102685259A CN201310268525A CN103324921A CN 103324921 A CN103324921 A CN 103324921A CN 2013102685259 A CN2013102685259 A CN 2013102685259A CN 201310268525 A CN201310268525 A CN 201310268525A CN 103324921 A CN103324921 A CN 103324921A
- Authority
- CN
- China
- Prior art keywords
- hand
- finger
- band
- image
- mobile identification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention provides a non-limited mobile identification method based on inner finger creases. The method utilizes a camera of mobile equipment as collecting equipment to collect hand images, can transmit collected samples to a server through a network, can automatically carry out hand area detection, inner finger crease area location and inner finger crease characteristic extraction, and can carry out characteristic comparison on inner finger crease samples in a database to achieve identity identification based on inner finger crease biological characteristics. The invention further provides mobile identification equipment of the mobile identification method based on the inner finger creases. The mobile identification equipment comprises a collecting module, a transmitting module, a preprocessing module, a processing module and a deciding module. The mobile identification equipment has the advantages of being initiatively applied to mobile environments, being adaptable to different background changes, illumination change, posture changes and sight point changes, having certain tolerance on dislocation, having high match success rate, being a convenient, efficient and reliable identity identification system and having good application prospect in the safety field.
Description
Technical field
The present invention relates to a kind of mobile recognition technology, particularly a kind of mobile identification method and mobile identification equipment thereof based on interior finger band.
Background technology
Along with popularizing of the mobile devices such as flat board, smart mobile phone, we need the work of strict authentication also more and more frequent at the affair of processor Migong, individual privacy affairs etc. on these equipment.But traditional passing through arranges the method that username and password is identified identity, do not have uniqueness, and easily passes out of mind, cracks or steal.From PIN(people's recognition code) different, biological characteristic can not pass into silence, and loses or steals, and can not be copied easily or share, and therefore the relatively traditional encryption recognition system of identification system take biological characteristic as the basis has very large advantage.
Now there have been many living creature characteristic recognition systems that are widely used, such as fingerprint, iris etc.Existing biometrics identification technology has following carrying a little:
The first, most biological identification technology depends on professional collecting device and builds a controlled sample collection environment to simplify recognizer, improves accuracy of identification.Even part adopts common camera, still there is certain requirement in environment, accuracy of identification is not high.Therefore, under mobile uncontrollable environment, prior art is difficult to obtain good recognition effect.
The second, based on the recognition methods of behavioural characteristic, gait for example, sound, signature and keystroke typewriting etc., accuracy of identification is not high and easily imitated.Therefore, for the hardware condition of existing mobile device, general consideration utilizes camera to gather the image informations such as palmmprint, interior finger band as basis of characterization.
Face recognition at Human and machine recognition of faces:a survey(people and machine: investigation), movable recognition system based on face has been proposed, but, this system needs a huge database that comprises the shooting sample under a large amount of different light and the posture condition in order to realize reliability.And for Palm Print Recognition System, they often need very high-resolution picture to carry out identification to obtain enough structural informations, yet when posture changed, high resolving power can cause again picture obviously to be subject to the impact of projective transformation simultaneously.
At Illumination ratio image:synthesizing and recognition with varying illuminations(illumination quotient images: variable light according under synthetic and identification) in mention, compare with sound, face and palmmprint, refer to band because its physical feature is fit to mobile biological recognition system more.Refer to band except smooth surface is arranged, the abundant structural information that is difficult in addition imitate is again because it is in a very little zone, so its projective transformation is very little even can ignore.That is to say, under the low resolution condition, utilize to refer to that band is identified and to obtain than using the higher accuracy rate of palmmprint identification.
Refer to that in the use that has proposed at present band carries out in the method for bio-identification, nearly all there is not the identification difficulty that fine solution posture changes and illumination condition brings, the picture that they mostly need the use specific installation and take under ecotopia, and do not consider the impact that when condition is inconsistent, causes.
Document A Biometric Identification System Based on Eigenpalm and Eigenfinger Features(refers to the living creature characteristic recognition system of feature based on the feature palm and feature) and A multi-matcher system based on knuckle-based features(based on many adaptations system of articulations digitorum manus feature) in the method that proposes need to use scanning device to obtain picture, document Online finger-knuckle-print verification for personal authentication(is for the online finger band authentication method of person identification) in the same system that need to be special of method.
The palm grain identification method of document Palmprint Recognition across Different Devices(on distinct device) although in method be to gather the palm picture with mobile device, but it requires picture background must be complete black, and illumination will be accomplished even as best one can.
Summary of the invention
Primary and foremost purpose of the present invention is to overcome the shortcoming of prior art with not enough, and a kind of mobile identification method based on interior finger band is provided, and the method has broken through in the existing method restriction of environmental baseline when gathering image, thereby is more applicable for mobile device.
Another object of the present invention is to overcome the shortcoming of prior art with not enough, a kind of mobile identification equipment of realizing based on the mobile identification method of interior finger band is provided, and this equipment mainly comprises the functions such as image acquisition, hand detection, area-of-interest (ROI) location, interior finger band feature extracting and matching.
Primary and foremost purpose of the present invention is achieved through the following technical solutions: a kind of mobile identification method based on interior finger band may further comprise the steps:
T1: gather image, the user adopts the hand images of the camera shooting that configures on the mobile device under mobile environment.
T2: hand detects, and after processing based on the method for the skin model of mixed Gauss model and morphologic expansion, corrosion, extracts complete and accurate hand profile from source images.
The T3:ROI location utilizes finger reference point localization method and Radon projection further to be partitioned into four area-of-interests from the hand images that T2 extracts.
T4: the feature extraction of interior finger band, extract the characteristic information in ROI zone based on the Competition coding method, comprise orientation map(directional diagram) and energy map(energygram).
T5: interior finger band characteristic matching, two feature map obtaining among the above-mentioned steps T4 and the characteristic in the server are mated and obtain final matching result based on the region histogram statistical method.
T2: hand detects and mainly comprises coarse localization (T21) and accurately locate (T22) two steps:
T21: coarse localization, as training set, utilize expectation maximization (EM) algorithm after improving with pre-prepd typical broca scale picture, come the iterative computation likelihood function by existing data, make it to converge on certain optimal value, thereby automatically obtain the parameter of Gaussian mixture model.Whether each pixel of judging the image that gathers meets certain gaussian kernel, then is judged as the hand skin zone if meet, otherwise is judged as the background area, thereby obtain rough hand images.
T22: accurately locate, utilize morphological method to process rough hand images obtained above, can remove duck eye and the noise of the inside, thereby obtain complete and accurate hand images and outline line.
T3: after extracting complete accurate hand images, from image, be partitioned into four area-of-interests.Main dividing for two steps (T31 and T32):
T31: the location of finger and cutting apart.The distance relation positioning datum point that we utilize outline line to put in hand images obtained above: five finger tip points and four finger valley points.Estimate the axis of every finger finger and the zone of extracting finger based on these reference points.Owing to referring in the thumb that the band inclusion information is less, so we only locate all the other four fingers.
T32: the finger areas image that T31 is cut apart rotates to level by the axis, and all samples are cut to unified size.Sample is done the Radon projection, obtain obvious two peak regions, be interior finger band region.The recycling low-pass filtering obtains the particular location coordinate, obtains final ROI zone.
T4: utilize the method based on Competition coding, from the region of interest area image that T3 obtains, extract the feature for identification.Method specifically describes: we utilize Gabor filtering to catch the directional information of the picture of publishing picture from image.After the Gabor filtering of obtaining different directions, the source images that utilizes the Gabor of each direction to check area-of-interest carries out convolution operation and finally obtains n opening the response image, the corresponding n of a difference different directions, the direction that comprises the minimum response value will be elected to be (being dominant) direction, and form the directional diagram (orientation map) that a width of cloth represents each pixel principal direction; Meanwhile, this minimum response can be stored into another width of cloth figure, through the negate of numerical value, the energygram (energy map) that quantization operation generates each pixel principal direction weights of description.The structure feature information of final orientation map and the common presentation video of energy map.
T5: the feature matching method based on the region histogram statistics mainly comprises two steps (T51 and T52):
T51: we are obtain orientation map(directional diagram from T4) and energy map(energygram) carry out the statistics with histogram of part.We are divided into many overlapping zones to source images, are called piece.Each piece is comprised of some nonoverlapping cells.Each cell corresponding a histogram that is formed by Nearest Neighbor with Weighted Voting by interior pixels.The voting results of each grid, namely corresponding histogram has formed the feature of each piece, and the feature of each piece has formed again whole Characteristic of Image vector, is used for final identification.
T52: the Euclidean distance that calculates the proper vector that stores in above-mentioned proper vector to be matched and the background data base.The present invention obtains a threshold value by the method for machine learning, if Euclidean distance less than threshold value, then characteristic matching success, the authentication of completing user identity.Owing to having adopted the method for statistics with histogram in the piece, matching algorithm of the present invention is very inresponsive to two position relationships of scheming between the piece to be matched within the specific limits, so when translation when dislocation that exists between the sample to be matched in the certain limit, still can effectively show the similarity degree of two figure, so the present invention has certain tolerance to dislocation.
Another object of the present invention is achieved through the following technical solutions: a kind of mobile identification equipment of realizing based on the mobile identification method of interior finger band mainly comprises with lower module:
Acquisition module, the hand images that is used for taking the user;
Transport module, the user's hand images that is used for photographing from described collecting unit by Internet Transmission is to server end;
Pretreatment module is used for detecting hand region from user's hand images of described server end, and refers to the band zone in the location;
Processing module is located the result who refers to the band zone based on described pretreatment unit, is used for extracting the also interior finger band information of match user;
Decision-making module based on the result of the interior finger band information of described processing unit match user, is used for determining the result of checking or identification.
The equipment that described acquisition module uses is the camera that is equipped with on the mobile device, and it is vertical that the direction of camera and the hand the five fingers launch formed plane, and the shake of camera is in 20 degree scopes.
Hand and background thereof had following characteristics when acquisition module gathered:
Background there are differences with the color of hand skin;
Bright and clear, there is not strong shadow;
The five fingers are naturally open and flat, separately;
Described pretreatment module comprises the positioning unit of hand region detecting unit and interior finger band, described hand region detecting unit adopts the skin model based on mixed Gauss model to extract hand region, described interior finger band positioning unit utilizes the geological information of hand profile to extract finger areas, finger areas is done the Radon projection, refer to the band zone in extracting.
Described processing module comprises feature extraction unit and the matching unit in interior finger band zone,
The extraction unit of described feature adopts the character representation method of described Competition coding; The unit is input as described area-of-interest zone, is output as directional diagram claimed in claim 3 and energygram;
The matching unit of described feature adopts described method based on partial statistics; The unit is input as described directional diagram and energygram, and output is the Euclidean distance of Characteristic of Image vector in this input picture and the database that mates with it.
Described decision-making module is based on the matching result of one whole hand and does decision-making, and described Synthetic Decision Method based on multizone comprises two kinds of patterns: Validation Mode and recognition mode;
Described Validation Mode is to carry out man-to-man coupling, judges whether image is the same hand in target image and the database;
Described recognition mode is the coupling of one-to-many, finds the hand that mates most from the database of hand, if the decision-making energy value is lower than threshold value, refers to not exist in the band database picture of the user's who detects hand in then judging.Here adopted the method for weighted mean value to calculate the decision-making energy value, because the common inclusion information of little finger band is less, so weights are lower, the weights of forefinger, middle finger, the third finger, little finger of toe are set as respectively 0.3,0.3,0.3,0.1, therefore, total decision-making energy value T=0.3* (T2+T3+T4)+0.1*T5, wherein T2, T3, T4, T5 represent respectively the energy value of forefinger, middle finger, the third finger, little finger of toe.
Principle of work of the present invention: the present invention utilizes the camera collection hand images of mobile device, utilization is extracted hand region based on the skin model of mixed Gauss model, by referring to the zone of band in the geometric properties location of analyzing the hand outline line, utilization refers to the band feature and adopts region histogram to mate in extracting based on the Gabor filtering technique of Competition coding, thereby realized the identification based on interior finger band biological characteristic.
The present invention has following advantage and effect with respect to prior art:
1, this method is by utilizing comparatively stable significantly interior band and the image processing method of referring to of feature, effectively reduce illumination and user's posture to the impact of recognition image, overcome the face recognition at document Human and machine recognition of faces:a survey(people and machine: need a shortcoming that comprises the huge database of the shooting sample under a large amount of different light and the posture condition investigation).
2, at document Illumination ratio image:synthesizing and recognition with varying illuminations(illumination quotient images: mention synthetic and identification under variable light is shone), compare with sound, face and palmmprint, refer to band because its physical feature is fit to mobile biological recognition system more.Refer to band except smooth surface is arranged, the abundant structural information that is difficult in addition imitate is again because it is in a very little zone, so its projective transformation is very little even can ignore.That is to say, this method utilization refers to that band is identified and can obtain than using the higher accuracy rate of palmmprint identification under the low resolution condition.
3, this method only need to can realize with the mobile device of common camera, overcome document ABiometric Identification System Based on Eigenpalm and Eigenfinger Features(refers to feature based on the feature palm and feature living creature characteristic recognition system), A multi-matcher system based on knuckle-based features(is based on many adaptations system of articulations digitorum manus feature) and document Online finger-knuckle-print verification for personal authentication(for the online finger band authentication method of person identification) in need the additional limits of specific installation.
4, with the palm grain identification method of document Palmprint Recognition across Different Devices(on distinct device) in to require picture background must be complete black, and illumination will accomplish that as best one can method is compared uniformly, this method is utilized significantly interior band and the image processing method of referring to of feature, effectively reduces the requirement of picture background and photoenvironment.
5, compared with prior art, the advantage of this method is to reduce largely the condition restriction when gathering image.Traditional method must be carried out the collection of recognition image by specific auxiliary sampling instrument, or requires stricter to surrounding environment when gathering image.And our method refers to the band feature in having utilized comparatively significantly, only need to be with the camera that configures on the mobile device as sampling instrument.In the process of extracting feature, reduce as far as possible again image by the whole bag of tricks owing to the exposure under the different situations, rotation, the mobile variation that causes, thereby different environmental baselines is had stronger robustness.
Description of drawings
Fig. 1 is based on the skin model of mixed Gauss model.
Fig. 2 is the workflow diagram of on-site identification process of the present invention.
Fig. 3 a obtains the wide image of coarse handwheel after detecting through skin model respectively.
Fig. 3 b is the wide image of comparatively complete accurate handwheel that obtains after learning processing through overexpansion, etch state.
Fig. 4 a is finger reference point location schematic diagram.
Fig. 4 b is the schematic diagram of location, finger axis.
Fig. 4 c is the schematic diagram after finger is cut apart.
Fig. 5 is based on the schematic diagram of area-of-interest (ROI zone) location of the interior finger band of RADON projection.
Fig. 6 a is the method schematic diagram with localized mass sector scanning directional diagram.
Fig. 6 b is the block structure schematic diagram.
Fig. 6 c is the bicubic interpolation schematic diagram.
Fig. 7 is equipment structure chart of the present invention.
Fig. 8 is the process flow diagram of hand images pretreatment module.
Fig. 9 is based on the ROC figure of ecotopia and mobile environment.
Embodiment
The present invention is described in further detail below in conjunction with embodiment and accompanying drawing, but embodiments of the present invention are not limited to this.
Embodiment
Whole identifying of the present invention comprises two parts: preliminary preparation and field conduct.
Preliminary preparation comprises two parts: set up hand images database and skin model.
Set up the hand images database, gather targeted customer's hand images.Collecting device is common camera, and its acquisition condition mainly comprises following characteristics: background there are differences with the color of hand skin; Light is relatively more sufficient, does not have strong shadow; The five fingers are naturally open and flat, separately; The direction of camera is over against the plane of hand, and shake is in 20 degree scopes.
Set up skin model, the extraction typical case area of skin color from the hand images database utilizes the expectation-maximization algorithm after improving as training set, automatically obtains the parameter of Gaussian mixture model.As shown in Figure 1, shown the distribution of 8 gaussian kernel that represent skin model.
The flow process of field conduct is as shown in Figure 2:
#1 gathers user's hand images;
#2 arrives server with image uploading;
The #3 hand extracts;
#4 points orientation and segmentation;
The area-of-interest that refers to band in the #5 location;
Refer to the band feature extraction in the #6;
#7 mates the feature that extracts in database;
The #8 decision making package;
The #9 server returns matching result to the mobile terminal;
The ins and outs in concrete each step are as follows in the process flow diagram:
#1 gathers user's hand images: the user uses the camera that configures on the mobile device to take hand images, and the characteristics of its acquisition condition are as follows: background there are differences with the color of hand skin; Light is relatively more sufficient, does not have strong shadow; The five fingers are naturally open and flat, separately; The direction of camera is over against the plane of hand, and shake is in 20 degree scopes; Make palm comprise that finger occupies whole image frame substantially.
#2 arrives server with image uploading: mobile device uploads to background server by 3G or this WIFI network insertion internet with the hand images that collects.
The #3 hand extracts: the coarse localization of carrying out hand region based on the skin model that obtains in advance, specific descriptions are: judge whether each pixel meets certain gaussian kernel on the image that gathers, if meet and then be judged as the hand skin zone, otherwise be judged as the background area, thereby obtain rough hand images, shown in Fig. 3 a.On this basis, utilize morphological method to remove duck eye and noise in the above-mentioned rough result, thereby obtain complete and accurate hand images and outline line, shown in Fig. 3 b.
#4 points orientation and segmentation: based on hand region obtained above, we utilize the distance relation positioning datum point of putting on the outline line: five finger tip points and four finger valley points.Shown in Fig. 4 a, we are since an end points P, along the distance of outline line pointwise statistics with some P, when variable in distance trend occurs (for example increasing to the T5 front distance from P when obviously changing always, behind T5, begin to reduce), choose currently as unique point, we can obtain five finger tip points of T1-5 like this, and B1, B3, B4, B5 refer to a valley point.Moreover, our outline line between T2 and B1 finds the some B2 nearest with B3, in like manner finds a B6 as the New Characteristics point, and we have just determined the Position Approximate of finger substantially like this.
Take middle finger as example, shown in Fig. 4 b, we obtain M1, M2 ', M3, four trisection points of M4 ' with B3T3, two curve trisections of B4T3, and then beginning up and down from the M2 point, both direction travels through each point along outline line, calculate the distance of this point and M1, obtain 1 nearest M2, in like manner obtain M4, connect M1M2, M3M4, get the line D2V2 of these two sections emphasis, be the axis of this finger.Do rectangle frame take this axis as axle, can obtain finger areas.
Because refer in the thumb that the band inclusion information is less, we only locate all the other four fingers, shown in Fig. 4 c.
The area-of-interest that refers to band in the #5 location: the finger areas that previous step splits is rotated to level, each finger areas is done the Radon projection, each finger areas can obtain two peak values, utilize low-pass filtering to obtain the interior finger band zone of determining to each finger areas, Fig. 5 represents to refer in the interior place of a finger areas extraction of band.
Refer to the band feature extraction in the #6: extracting method in the past change color smoother situation under can lose efficacy, in order to address this problem, the present invention adopt the Competition coding method from the ROI area image internally finger band structural information extract.In the present invention, the Gabor wave filter is used to extract the directional information that refers to band.The present invention carries out process of convolution to image pixel by the Gabor kernel on n angle, convolution kernel draws by a Gabor function calculation based on neuro-physiology.This function is:
x′＝cosθ·x+sinθ·y，
Wherein θ is little wave line of propagation, generally gets 18, and namely per 10 degree are got one and represented direction; σ is the standard deviation of Gaussian profile; λ and
Respectively frequency and the phase place of sine function.Through obtaining 18 response images after the convolution, then contrast the sensitivity of each computing unit on 18 directions, response is less, the expression directivity is stronger, the direction of response minimum is made as the principal direction of this computing unit, the principal direction of all computing units is stored, form the directional diagram (orientation map) of each picture element principal direction on the Description Image.Meanwhile, this minimum response can be stored into another width of cloth figure, deducts minimum value by the data to this width of cloth figure the inside, get inverse, and quantize to operations such as [0,1.0], generate the energygram (energy map) of the weights that represent each pixel principal direction.The weights of energy map are larger, represent this pixel just stronger at the lines that this principal direction embodies.The structure feature information of final orientation map and the common Description Image of energy map.
#7 mates the feature that extracts in database: structural information (the orientation map that has obtained interior finger band, energy map) after, the present invention adopts the method for region histogram statistics to realize the image spatial feature coupling of tolerance dislocation.Specifically can be described as:
The first, define a piece zone, 2x2 subelement of each piece district inclusion, each subelement comprises again 4X4 pixel, shown in Fig. 6 b.
The second, from left to right, from top to bottom, with a piece sector scanning right 6 described orientation map(directional diagrams).From left to right mobile step-length is 4 pixels, and mobile step-length is 4 pixels from top to bottom, shown in Fig. 6 a of institute.
The 3rd, in each piece zone, the zone of the orientation map (directional diagram) of correspondence is carried out based on histogrammic statistics.For each subelement, add up respectively energy value separately on 6 directions.When calculating the contribution of each pixel to 6 different directions of different subelements, must carry out bicubic interpolation to weights according to locus and the orientation angle of this pixel, shown in Fig. 6 c.The weights here refer to the value of this pixel correspondence on energy map (energygram).Thereby each subelement comprises the proper vector of one 6 dimension, also means, each piece comprises the proper vector of one 24 dimension.The proper vector in each piece zone is combined, has just formed the proper vector of describing entire image.
The 4th, calculate two images similarity structurally, calculate exactly the distance of two width of cloth image feature vectors, adopt the Euclidean distance computing method.By a large amount of experimental datas, whether we have predicted one reliably apart from as threshold value, mate based on these threshold decision two interior finger bands.Threshold value described here is an empirical value of testing based on great many of experiments.
#8 decision making package: in order to obtain more reliable matching structure, the present invention adopts the Synthetic Decision Method based on multizone, specifically can be described as: for each hand, the zone of coupling comprises the interior finger band of the 3rd joint of the four fingers finger except thumb, that is to say, a handbag contains the subregion of four couplings.As long as the matching result of four sub regions exists two or more results to be judged as coupling, think that then the image of corresponding two width of cloth hands is from the same hand.The decision-making energy value is the weighted mean of the coupling energy value in the interior finger band zone on the coupling.
The #9 server returns matching result to the mobile terminal: the result that server will mate passes to corresponding mobile terminal by network, and matching result shows in the mobile terminal.
Based on above flow process, what propose among the present invention comprises 5 modules based on the required equipment of the unrestricted mobile identification method of interior finger band, as shown in Figure 7.Equipment is comprised of acquisition module, transport module, pretreatment module, processing module and decision-making module.
Acquisition module is responsible for taking user's hand images;
User's hand images that transport module is responsible for photographing from described collecting unit by Internet Transmission is to server end;
Pretreatment module is responsible for detecting hand region from gather next image, and refers to band ROI zone in the location;
Processing module is responsible for feature extraction and matching is carried out in the interior finger band ROI zone that obtains;
Decision-making module is responsible for determining based on matched data the result of checking or identification.
Pretreatment module comprises the positioning unit of hand region detecting unit and interior finger band, it is characterized in that: described hand region detecting unit adopts the skin model based on mixed Gauss model to extract hand region, described interior finger band positioning unit utilizes the geological information of hand profile to extract finger areas, finger areas is done the Radon projection, refer to band ROI zone in extracting, idiographic flow as shown in Figure 8.
Processing module comprises feature extraction unit and the matching unit in interior finger band zone, wherein, feature extraction unit adopts the character representation method of Competition coding, is input as interior finger band ROI area image, is output as directional diagram (orientation map) and energygram (energy map); The characteristic matching unit adopts the method based on partial statistics, is input as directional diagram (orientation map) and energygram (energy map), and output is the Euclidean distance of Characteristic of Image vector in target image and the database.
Decision-making module comprises checking (verification) unit and identification (identification) unit.Decision-making module is based on the Comprehensive Model of multizone, and each hand coupling 4 each ROI zone is respectively the third knuckle of middle forefinger, middle finger, the third finger, little finger of toe.Authentication unit is man-to-man coupling, judges whether image is the same hand in target image and the database; Recognition unit is the coupling of one-to-many, finds the hand of coupling from the database of hand, and the highest hand of energy value of namely making a strategic decision if the decision-making energy value is lower than threshold value, is then thought the picture that does not have the same hand in the database.
In order to verify method of the present invention, we respectively under the consistent ecotopia of illumination and under mobile environment, obtain sample and carry out Matching Experiment, as table 1 and shown in Figure 9 be experimental result of the present invention:
| Sample type | Sample size | Correct coupling | Erroneous matching | Be matched to power |
| Ecotopia | 1000 | 992 | 8 | 99.2% |
| Mobile environment | 500 | 485 | 15 | 97.0% |
Table 1
Under the mobile environment, 22-〉15,
Can find out from experimental result, the identification of the present invention under mobile environment can also keep preferably matching effect, so the method before comparing, our invention is more suitable for being applied in the mobile device.
Fig. 9 represents is the present invention under the desirable environment and the experimenter's performance curve under the mobile environment, wherein GAR represents correct receptance, FAR represents false acceptance rate, as can be seen from the figure, our method can be in the situation that very low FAR under ecotopia, obtain very high GAR value, the comprehensive matching rate surpasses 99%.Even under mobile environment, we also can easily reach 97% the power that is matched to, and this just shows that performance of the present invention satisfies user demand.
Above-described embodiment is the better embodiment of the present invention; but embodiments of the present invention are not restricted to the described embodiments; other any do not deviate from change, the modification done under Spirit Essence of the present invention and the principle, substitutes, combination, simplify; all should be the substitute mode of equivalence, be included within protection scope of the present invention.
Claims (10)
1. the mobile identification method based on interior finger band is characterized in that, may further comprise the steps:
Step 1, set up with indoor finger band database;
Step 2, automatically detect hand region;
Refer to the band zone in step 3, the location;
Step 4, the character representation that illumination, attitude and viewpoint is had robustness;
Step 5, dislocation had the characteristic matching of tolerance;
Step 6, based on the multizone decision making package.
2. the mobile identification method based on interior finger band according to claim 1 is characterized in that,
In the described step 1, the indoor finger band of described usefulness database adopts the camera on the mobile device to take the hand images that the user points when mobile and sets up;
In the described step 3, refer to the method in band zone in the described location, may further comprise the steps:
The first step: set up the mixed Gauss model that is generated as training set by the broca scale picture, improve expectation-maximization algorithm and obtain the parameter of each Gaussian distribution in the described mixed Gauss model, and set up skin model;
Second step: in each Gaussian distribution of the skin model that each the pixel color value substitution of the hand images of described shooting has been set up, if pixel value meets Gaussian distribution, then differentiate and be hand region, otherwise be the background area;
The 3rd step: with the method for the dilation and erosion in the morphology image of described shooting is carried out filtering, eliminate scrappy zone, obtain complete hand profile;
The 4th step: with the distance relation positioning datum point of putting on the described complete hand outline line;
The 5th step: position and the axis of determining every finger in the image of described shooting with the geometric relationship of described reference point and described outline line;
The 6th step: use the image of the described shooting of analysis of Radon projection to obtain area-of-interest.
3. the mobile identification method based on interior finger band according to claim 1, it is characterized in that, in the described step 4, the described character representation that illumination, attitude and viewpoint are had a robustness refers to the character representation method based on Competition coding, and described character representation method based on Competition coding may further comprise the steps:
A, for each pixel in interior finger band zone, calculate the energy value of the Gabor filtering of n direction, that direction of energy value minimum is made as the principal direction of this pixel; A described n direction refers to 360 degree are divided into n interval, and n is the integer greater than 18;
Refer to band image process Competition coding in B, the width of cloth, export the characteristic image of the same resolution of two width of cloth; Represent the image of each pixel principal direction, be called directional diagram; Represent the weights of each pixel principal direction, be called energygram.
4. the mobile identification method based on interior finger band according to claim 1, it is characterized in that, in the described step 5, the described feature matching method that dislocation is had a tolerance refers to the feature matching method based on partial statistics, and described feature matching method based on partial statistics may further comprise the steps:
⑴ define a piece zone, comprises square subelement of c, and each subelement comprises again square pixel of S;
⑵ from left to right, from top to bottom, scan with a piece zone on described directional diagram; From left to right mobile step-length is d1, and mobile step-length is d2 from top to bottom, and step-length d1 and d2 must guarantee between the piece zone that at least 1/3 zone has overlapping; Each time scanning is carried out based on histogrammic statistics space corresponding to piece zone; Histogram for the piece zone is added up;
In the described step (2), described statistical method of adding up for the histogram in piece zone is:
1. each subelement is added up respectively m the energy value on the direction, m must be less than n claimed in claim 3;
When 2. calculating each pixel to the contribution of m different directions of different subelements, according to locus and the orientation angle of this pixel weights are carried out bicubic interpolation, described weights are the value of this pixel correspondence on energygram;
3. the proper vector dimension of each subelement is n, total c unit, then the proper vector dimension in each piece zone be c square with the product of n;
4. the proper vector in each piece zone is combined, and forms the proper vector of describing entire image;
5. adopt the Euclidean distance computing method to calculate the distance of image feature vector in target image and the database, calculate structural similarity between the image to be matched.
5. the mobile identification method based on interior finger band according to claim 1 is characterized in that, in the described step 6, the method for described decision making package based on multizone may further comprise the steps:
The interior finger band of I, the forefinger of setting detected hand, middle finger, the third finger, these four fingers of little finger of toe is as the subregion of coupling;
II, every sub regions is carried out described character representation based on Competition coding and separately based on the characteristic matching of partial statistics;
The subregion of these four fingers of III, forefinger, middle finger, the third finger and little finger of toe with compare with the data of indoor finger band database, if exist finger more than two or two when being complementary with the data of indoor finger band database, judge that then the image of the hand that mates in the image of the hand that is detected and the database is complementary;
The weighted mean value of the coupling energy value in IV, the decision-making energy value interior finger band zone that to be these four fingers of forefinger, middle finger, the third finger and little finger of toe be complementary with data with indoor finger band database.
6. a realization is characterized in that based on the mobile identification equipment of the mobile identification method of interior finger band, comprising:
Acquisition module, the hand images that is used for taking the user;
Transport module, the user's hand images that is used for photographing from described collecting unit by Internet Transmission is to server end;
Pretreatment module is used for detecting hand region from user's hand images of described server end, and refers to the band zone in the location;
Processing module is located the result who refers to the band zone based on described pretreatment unit, is used for extracting the also interior finger band information of match user;
Decision-making module based on the result of the interior finger band information of described processing unit match user, is used for determining the result of checking or identification.
7. realization according to claim 6 is based on the mobile identification equipment of the mobile identification method of interior finger band, it is characterized in that: the equipment that described acquisition module uses is the camera that is equipped with on the mobile device, and it is vertical that the direction of described camera and the five fingers of hand launch formed plane.
8. realization according to claim 6 is based on the mobile identification equipment of the mobile identification method of interior finger band, it is characterized in that, described pretreatment module comprises the positioning unit of hand region detecting unit and interior finger band, described hand region detecting unit adopts the skin model based on mixed Gauss model to extract hand region, described interior finger band positioning unit utilizes the geological information of hand profile to extract finger areas, finger areas is done the Radon projection, refer to the band zone in extracting.
9. realization according to claim 6 is characterized in that based on the mobile identification equipment of the mobile identification method of interior finger band: described processing module comprises feature extraction unit and the matching unit in interior finger band zone,
The extraction unit of described feature adopts the character representation method of described Competition coding; The unit is input as described area-of-interest, is output as directional diagram claimed in claim 3 and energygram;
The matching unit of described feature adopts described method based on partial statistics; The unit is input as described directional diagram and energygram, and output is the Euclidean distance of Characteristic of Image vector in this input picture and the database that mates with it.
10. realization according to claim 6 is based on the mobile identification equipment of the mobile identification method of interior finger band, it is characterized in that: described decision-making module is based on the matching result of hand and does decision-making, described Synthetic Decision Method based on multizone comprises two kinds of patterns: Validation Mode and recognition mode;
Described Validation Mode is to carry out man-to-man coupling, judges whether the image of detected hand and the image of the hand in the interior finger band database are the same hand that is:;
Described recognition mode is the coupling of one-to-many, finds the image of the hand that mates most from the interior finger band database of hand that is:, if the decision-making energy value is lower than threshold value, refers to not exist in the band database picture of the user's who detects hand in then judging.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310268525.9A CN103324921B (en) | 2013-06-28 | 2013-06-28 | A kind of mobile identification method based on interior finger band and mobile identification equipment thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310268525.9A CN103324921B (en) | 2013-06-28 | 2013-06-28 | A kind of mobile identification method based on interior finger band and mobile identification equipment thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103324921A true CN103324921A (en) | 2013-09-25 |
| CN103324921B CN103324921B (en) | 2016-12-28 |
Family
ID=49193652
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310268525.9A Active CN103324921B (en) | 2013-06-28 | 2013-06-28 | A kind of mobile identification method based on interior finger band and mobile identification equipment thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103324921B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077560A (en) * | 2014-01-13 | 2014-10-01 | 北京市公安局刑事侦查总队 | Fingerprint comparison method |
| CN104615635A (en) * | 2014-11-10 | 2015-05-13 | 南方医科大学 | Direction feature based palm vein category index establishment method |
| CN104820828A (en) * | 2015-04-30 | 2015-08-05 | 武汉大学 | Identity authentication method based on multi-directional mixed features of dual inner phalangeal prints |
| WO2017000342A1 (en) * | 2015-06-29 | 2017-01-05 | 宇龙计算机通信科技(深圳)有限公司 | Fingerprint recognition method and terminal |
| CN107146217A (en) * | 2017-04-07 | 2017-09-08 | 北京工业大学 | A kind of image detecting method and device |
| CN110728232A (en) * | 2019-10-10 | 2020-01-24 | 清华大学深圳国际研究生院 | Hand region-of-interest acquisition method and hand pattern recognition method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101504722A (en) * | 2009-03-20 | 2009-08-12 | 公安部第一研究所 | Multi-mode combined identity authentication system and method |
| US20100119125A1 (en) * | 2008-11-12 | 2010-05-13 | Chi Mei Communication Systems, Inc. | Fingerprint identification system and method of an electronic device |
-
2013
- 2013-06-28 CN CN201310268525.9A patent/CN103324921B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100119125A1 (en) * | 2008-11-12 | 2010-05-13 | Chi Mei Communication Systems, Inc. | Fingerprint identification system and method of an electronic device |
| CN101504722A (en) * | 2009-03-20 | 2009-08-12 | 公安部第一研究所 | Multi-mode combined identity authentication system and method |
Non-Patent Citations (2)
| Title |
|---|
| LIN ZHANG 等: "Online finger-knuckle-print verification for personal authentication", 《PATTERN RECOGNITION》, vol. 43, no. 7, 27 January 2010 (2010-01-27), pages 2560 - 2571 * |
| 毛贤光: "指横纹识别系统及关键技术的研究与实现", 《中国博士学位论文全文数据库信息科技辑》, 15 December 2010 (2010-12-15) * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077560A (en) * | 2014-01-13 | 2014-10-01 | 北京市公安局刑事侦查总队 | Fingerprint comparison method |
| CN104077560B (en) * | 2014-01-13 | 2017-07-04 | 北京市公安局刑事侦查总队 | Fingerprint comparison method |
| CN104615635A (en) * | 2014-11-10 | 2015-05-13 | 南方医科大学 | Direction feature based palm vein category index establishment method |
| CN104615635B (en) * | 2014-11-10 | 2018-06-05 | 南方医科大学 | Palm vein classified index construction method based on direction character |
| CN104820828A (en) * | 2015-04-30 | 2015-08-05 | 武汉大学 | Identity authentication method based on multi-directional mixed features of dual inner phalangeal prints |
| CN104820828B (en) * | 2015-04-30 | 2018-05-11 | 武汉大学 | A kind of identity identifying method based on double interior multi-direction composite characters of phalangeal configurations |
| WO2017000342A1 (en) * | 2015-06-29 | 2017-01-05 | 宇龙计算机通信科技(深圳)有限公司 | Fingerprint recognition method and terminal |
| CN107146217A (en) * | 2017-04-07 | 2017-09-08 | 北京工业大学 | A kind of image detecting method and device |
| CN110728232A (en) * | 2019-10-10 | 2020-01-24 | 清华大学深圳国际研究生院 | Hand region-of-interest acquisition method and hand pattern recognition method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103324921B (en) | 2016-12-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101763507B (en) | Face recognition method and face recognition system | |
| Yang et al. | A fingerprint verification algorithm using tessellated invariant moment features | |
| Raghavendra et al. | Scaling-robust fingerprint verification with smartphone camera in real-life scenarios | |
| Si et al. | Dense registration of fingerprints | |
| Feng et al. | Robust and efficient algorithms for separating latent overlapped fingerprints | |
| CN103324921A (en) | Mobile identification method based on inner finger creases and mobile identification equipment thereof | |
| CN102270308B (en) | Facial feature location method based on five sense organs related AAM (Active Appearance Model) | |
| Barpanda et al. | Iris feature extraction through wavelet mel-frequency cepstrum coefficients | |
| CN104200239A (en) | Image feature fusion identification based signature authentic identification system and method | |
| CN101558431A (en) | Face authentication device | |
| CN103268497A (en) | Gesture detecting method for human face and application of gesture detecting method in human face identification | |
| CN107169479A (en) | Intelligent mobile equipment sensitive data means of defence based on fingerprint authentication | |
| Labati et al. | Touchless fingerprint biometrics | |
| CN102254165A (en) | Hand back vein identification method based on fusion of structural coding features and texture coding features | |
| Krishneswari et al. | A review on palm print verification system | |
| US10121054B2 (en) | Systems and methods for improving spoof detection based on matcher alignment information | |
| Naderi et al. | Fusing iris, palmprint and fingerprint in a multi-biometric recognition system | |
| Zhou et al. | Double biologically inspired transform network for robust palmprint recognition | |
| CN110008825A (en) | Palm grain identification method, device, computer equipment and storage medium | |
| Bera et al. | Finger contour profile based hand biometric recognition | |
| Tarar et al. | Fingerprint image enhancement: iterative fast Fourier transform algorithm and performance evaluation | |
| Lin et al. | A combination recognition method of palmprint and palm vein based on gray surface matching | |
| US10176362B1 (en) | Systems and methods for a gradient-based metric for spoof detection | |
| Jain | Automatic Fingerprint Matching Using Extended Feature Set | |
| Naser et al. | Artificial Neural Network-Based Fingerprint Classification and Recognition. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |